Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Richard Koestner
McGill Univervitv
Richard M. Ryan
University,of Rochester
The finding that exlrinsic rewvards can undernineintrinsic motivation has been
highly controveisialsin.e itfirstoppeared(Deci, 1971). A meta-analysis pub-
lished in this journal (Cameron & Pierce 1994) conc:luded thlat tlhe underiniing effect was minimal and largely inmonsequentiialforeducationalpolicy.
However, a more recent meta-analysis (D)eci, Koestner, & Ryan, 1999)
showed that the Cameron andPie ce meta-analysiswas seriousvh flawed and
that its conclusions i-ere incorrect. This article briefly r eviews the results of
the more recent meta-analysis,which showed that tangible rewardsdo indeed
have a substantial undermining effrct. The meta-analysis provided strong
supportforcognitive evaluation theorv (Deci & Ryan, 1980), vhich Carneron
and Pierce had advocated abandoning.The results aie briefly discussed in
terms of their relevance for educationalpractice.
Gold stars, best-student awards, honor roles, pizzas for reading, and other
reward-focused incentive systems have long been part of the cu-rreticy of schools.
Typically intended to motivate or reinforce student learning, such techniques have
been widely advocated by some educators, although, in recent years, a few cominentators have questioned their widespread use. The coitroversy has been
prompted in part by psychological research that has demonstrated negative effects
of extrinsic rewards on students' intrinsic motivation to learn. Some studies have
suggested that, rather than always being positive niotivators, rewards can at tines
undermine rather than enhance self-motivation, ctuiosi-ty, interest, and persistence
at learning tasks. Because of the widespread ose of rewards in schools, a carefui
summary of reward effects on intrinsic motivation would seemt to be of consi(lerable importance for educators.
Accordingly, in the Fall 1994 issue of Review oaf Fducational Research,
Cameron and Pierce (1994) presented a meta-analysis of extrinsic reward effects
on ir.trinsic motivation, concluding that, overall, rewards do not decrease mttrinsic
motivation. Implicitly acknowledging that intrinsic motivation is important foi
learning and adjustment in educational setitings (see. e.g., Ryan & La Guardia,
1999), Cameron and Pierce nonetheless stated that "teachers have no reason to
I
resist ~imiplementi.ngilcentive systems in the classroom" (p. 397). Thev also advocated abandoning Deci and Ryan's (1980) cogniitive evaluation tleory (CET),
which had initially been lormulated to explain both positive and negative reward
effects on irtrinsic motivation.
In the Spring 1996 issue of RER, three comlmenitaries were published (Kohn,
1996: Lepper, Keavney, & Drake, 1996: Ryan & Dcci, 1996) arguing that Cameron
and Pierce's meta-analysis was flawed anid that its conclusions were unwarranted.
In that sanie issue, C.ameron and Pierce (1996) responded to the comnIentaries by
claiming that, rather than reanalyzing the data, the authors of the three conmmentaries had suggested "that the findings are invalid due to intentional bias, deliberate misrepresentation, and inept analysis" (p. 39). Subtitling their response "Protests
and Accusations Do Not Alter the Results," Cameron and Pierce stated that aniy
meaningful c iticism of their aricle would have to include a reanalysis of the data.
Subsequent to that interchange, Fisenberger and Cameron (1996) published an article in the American Psychologist summarizing the Cameronl and Pierce (1994)
nmeta-analysis and claiming that the so-cal'ed undermining of intrinsic motivation
by extrinsic rewards, which they said had become accepted as reality, was in fact
largely a myth.
We do not claim that there was "intentional bias" or "deliberate misrepresentatxon" in either the Cameron and Pierce (1994) meta-analysis or the Eisenberger and
Camneron (1996) article, but we do believe, as Ryan and Deci argued in 1996, that
Cameron and Plierce used somiC inappropriate procedures and made numerous
ernrs in tneir meta-analysis. 'herefore, because we believe the problems with their
meta-anaivsis made their conclusions invalid, because we agree that a useful critique
of their article nmust involve reanalysis of the data, and because the issue of reward
effcets on intrinsic motivation is extremely important for educators, we performed
a new meta-analysis of reward effects on intrinsic motivation (Deci, Koester, &
Ryan, 1999). Our Ineta-analysis included 128 experiments, organized so as to
provide a test of CET, much as Cameron and Pierce had done. The new mnetaanalysis, which we summarize in this article, showed that, in fact, tangible rewards
do significantlv and substantiallv undermineinthinsic motivation. The meta-analysis
provided strong support ior CER' and made clear that there is indeed reason for
teachers to exercise great care when using reward-based incentive systems.
The new meta-analysis was published in Psychological Bulletin (Deci ct al,
1999). Included in that article was an appendix table (here reproduced with permission as Table I a) listing every study in the meta-analysis and explaining exactly
where enors were made by Cameron and Picrce, how our meta-analysis corrected
their errors, and what studies were included in ours that had been overlooked or
onhitted bv them. The table allows interested readers to see for themselves exactly
how it is that Carmeron and Pierce's meta-analysis and our ineta-analysis anived at
such diff.erent conclusions.
In the seven years since the publication of Cameron and Pierce's (1994) article,
academics, school administrators, and classroom teachers from many countries
have spoken to uIs about the article, makihng it clear that the conclusions of the article had been widely disseminated and that the issue of reward effects is of considerable interest to educators around the world. Given the great importance of this
issue for education, then, the current article is intended to set the record straight for
the many readers of RER. In this article, we provide a brief description of CEr,
2
Aithough we do not usually use the term verbal rewards, preferring instead to
speak of "positive feedback," we do use that term here in order to include the
positive-feedback studies within the general category of reward effects. Verbal
rewards typically contain explicit positive perfornance feedback, so CET predicts
that they are likely to enhance perceived competence and thus enhance intrinisic
motivation. In the meta-analysis, we tested the hypothesis that verbal rewards
would enhance intrinsic motivation.
Nonetheless, verbal rewards can have a significant controlling aspect leading
peoplc to engage in behaviors specifically go gain praise, so verbal rewards have
3
the potential to undermine intrinsic motivation. The theory therefore suggests that
the interpersonal context within wlich positive feedback is adrninistered can influence whether it wili be interpreted as informational or controlling. As used here,
the term interpersonalcontext refers to the social amnbience of settings, such as
classrooms, as they influence people's experience of self-detenninaton (I)eci &
Ryani, 199 1). When studied in laboratory experinmets, thc interpersonal climate is
usually manipulated in terms of the interpersonal style used by the experimenter
when providing the feedback (e.g., Ryan, 1982; Ryan, Miins, & Koestner, 1983).
An interpersonal context is considered controlling to the extent that people feel
pressured by it to think, feel, or behave in particular ways. Verbal rewards administered within such a context are thus more likely to be experienced as controlling
rather than informational. For examiple, CFT suggests that if a teacher uses an
interpersonal style intended to make students do what he or she wants them to, verbal rewards administered by that teacher are likely to be expenienced as controlling. ina supplemiiental meta-analysis involving five studies, we tested the prediction
that controlling positive feedbaci woul.d lead to less intrinsic motivation than inforinational positive feedback.
7anagibleRewvar&d
Unlike:verbal rewards, tangible rewards are freqsuently offered to people as an
inducement to engage in a behavior in which they nmight not otherwise engage.
Thus, according to CET, tangible rewards will tend to be experienced as controlling, and as a result they will tend to decrease intrinsic mrotivation. The metaanalysis tested the hypothesis that, overall, tangible rewards would decrease intrinsic
motivation.
rewards) and those thtat require engaging in thic activity but do not require complet
ing it (herein referred to as engagement-conpingentrewards). We (e.g., DCci & Ryan,
1985) have considered the completion-contingenit and engagement-conitingent
4
Although few studies have manipulated the interpersonal context of performancecontingent rewards, Ryan et al. (I 983) compared a perform.ance-contingent rewards
group in which the rewards were administered in a relatively controlling manner and
one in which they were administered in a relatively non-controlling manner. As predicted, the controlling administration of perforrmance-contingent rewards led to
underm-ining of intrinsic motivation relative to the noncontrolling administration. In
terms of education, this is a particularly important finding because it sugges'ts that
when rewards are used in the classroom, it is important that the climate of the classroom be supportive rather than controlling so that the students will be less likely to
experience the rewards as controlling.
Method
Our meta-analytic strategy (Deci et at, 1999) involved a hierarchical
approach in which the results of 128 experiments were examined in two separate
mneta-analyses. The first involved 101 of the studies that had used a free-choice
behavioral measure of intrinsic motivation, and the second involved 84 of the
st.udies that had used self-reported interest as a dependent variable. In a hierarchical meta-analysis, one begins with the most general categorv and reports the
composite effect size. If the set of effects is heterogeneous, then one proceeds to
differentiate the overall category into meaningful subcategories in an attermipt to
achieve homogeneitv of effects within the subcategories. Thus, in both inetaanalyses (i.c., with the two dependent measures), we began by calculating the
effects of all rewards on intrinsic motivation and then systermatically differentiated the reward conditions. Only after we had exhausted all possible moderator
variables did we discard outliers to create homogeneity within subcategories.
Using this approach, we ended up discarding only about 4% of the effects as outliers, whereas Cameron and Pierce (1994) had discarded approximately 20% of
the effects as outliers.
In the differentiation, studies were first separated into those that examined verbal
rewards versus those that examined tangible rewards. 'hen tangible rewards, which
have been extensively studied, were analyzed as follows. 1'he effects of rewards that
were unexpected versus expected were examined separately. Studies of expected
tangible rewards were then separated into four groups, depending on what the
rewards were contingenit upon. The groups were as follows: taisk noncontingent
(rewards that did not explicitly require working on a task), engagement contingent
(rewards that did require working on the task), completion contingent (rewards that
required finishing a task), and performance contingent (rewards contingent upon a
specified level of perfornance at a task). As described subsequently, because the
perfornnance-contingent reward effects on the free-choice measure were heterogeneous, that category was furdter differentiated. Finally, in categories in which the
effect sizes were heterogeneous after all theoretically based differentiations had
been comrpleted, we compared thc effects of the reward types on schoolchildren
versus college students, an issue that had not been considered previously but
enmerged from an inspection of the data anid seemed very important in terms of the
educational relevance of the results.
Inclusion criteria for studies that spanned tie period 1971 to 1996 were the following. First, because intrinsic motivation is pertinent to tasks that people experience as interesting and because the field of inquiry has always been defined in terms
6
of reward effects on intrinsic motivation for interesting tasks, w emcluded only sttldies or conditions within studies if the target task was at least moderately interesting
(i.e., if it either was not defined a priori as a boring task by the experimienter or did
not have a prereward interest rating below the midpoint of the scale). In contrast,
Cameron and Pierce ( 1994) had aggregated across boring and interesting tasks without even addressing the issue in their article. Second, the analyses included only
studies tlat assessed inttinsic motivation after the rewards had been clearly ternninated, because while the reward is in effect participants' behavior reflects a mnix of
intrinsic anid extrinsic motivation. Camneron and Pierce, however, included assessments which they called intrinsic motivation but which had been taken while tle
reward contingency was still in effect. 'Third, studies wer e included only if they had
an appropriate no-rewar(i control group. Cameron and Pierce had made numerous
comparisons based on questhinable selections of control groups, at times even using
inappropriate control groups when appropiiate ones were available.
In conductfing the mneta-analyses, we used Cohen's d as the measure of efiect
size. It reflects the difference between the means of two groups divided by the
pooled within-group standard deviations, adjusted for sample size (H1edges &
Olkin, 1985). The mean of the control group was subtracted from the mean of the
rewards group, so a negative d reflects an "undernmining effect," whereas a positive d reflects anI "enhancement effect."
Means, standard deviations, t tests, F"tests, anid sample sizes were used to calculate d values. For any study in which insufficient data were provided to calculate an effect size, we assigned ani effect of d = 0.00, anid we included those imputed
values in all analyses. All effect-size computations and summiarv analyses were
done with DSTAT (Johnson, 1993), a meta-analytic software program. Each calculation of a composite effect size is accompanied by a 95% confidence interval
(CI) (for additional methodological details, see Deci et a.u, 1999).
Results
Effects of All Rewards
because Camleron and Pier ce (1994) calculated the effect of all rewards on intrinsic motivation in their meta-analysis, we also calculated it for comnparative purposes. h'I'eeffect of all types of rewards across all relevant studies revealed
significant undermining for the f'ree-choice behavioral measure of intrinsic notivation (k= 101; d=-0.24; Cl =-0.29,--0(19)2 although the overall effect for the
self-report. measure was not significant. T'hese and other nmajor results are summarized in Table 1.
As already nicitioned, we expected that all rewards would not affect intrinsic
motivation in a uniform way, and thus we both expected and found that the set of
effects for the aI-rewards category was heterogeneous. Consequently, we proceeded with more differentiated analyses of specific types of rewards, based on
both theoretical and empirical considerations. We fiTst separated studies of verbal
rewards fromn those of tangible rewards.
TABLE I
Major results of Mhe meta-analysis of tize eJfets of extrinsic rewards onfree-choice
rnlrinsirmolivation and self-reported interest, shown as Coluen's composite d
''ith k effects included
Self-reported
interest
Free-choice
behavior
d
All rewards
Verbal rewards
College
Children
Tangible rewards
Unexpected
Expected
Task noncontingent
Engagement conitingent
College
Children
Complet.on contingent
Perfornance contingent
Maximal reward
Not maximuni reward
Positive feedback control
Negative fe~edback control
-(.24*
0.33*
0.43*
0.11
0.1
-0.36*
--0.14
-0.40*
- 0.21 *
-0.43'
-0.44*
-0.284'
-0.15*
-0.88*
-0.20*
-0.03
101
21
14"
7"
92
9.34'
9"
92
7"
55
12a
39"
19,
32
18"
6"
I(P
dI
0.04
0.31*
84
21"
-0.07*
0.05
-0.07*
0.21
-0.15*
70
5"
69
5
35a
-0.17'
--O.O1
13"
29"
3"
"'These categories were not futuher differentiated and are homogeneous. Some of the studies used to cdetermine the overall conmposite effect size (i.e., for all rewards) in each mietaa.alysis had nmultiple reward conditions, so the stnms of the numbers of effect sizs in the
nmost differentiated categories of each neta-analysis are greater tian the nunmbers in the
all-rewards category. There were I50 effect sizes inthe most differentiated categories for
thc free-choice analyses, of which 6 were removed as outliers, and there were 114 effect
sizes in the most differentiated catego:ies of the self-report analyses, of which 6 were
removed as outliers.
* Significant at p < .05 or greater.
8
Tangible Rewards
Next, we tested the CET prediction that, overall, tangible rewards (including
material rewards, such as money anid prizes, and symbolic rewards, such as trophies and good player awards) would decrease intrinsic motivation, because tangible rewards are frequently used to persuade people to do things they wouid not
otherwise do, th-at is, to control their behavior. The meta-analysis included 92 tangible reward studies with a free-choice measure and 70 with a self-report measure.
As predicted by GET, results indicated that, on average, tangible rewards significantly underminied both free-choice intrinsic motivation (d= -.34; Cl -0.39, -0.28)
aid self-reported interest (d=-0.07; Cl =--0.13, -0.01). Of course, we have regularly arguedthat afill understandingof the effects of tmgiblerewarrds requires aconsideration of additional factors such as reward contingencv and interpersonal context,
but thcse results do highlight the general risks associated with the use of tangible
rewards as a motivator.
Because age cffects had emerged for verbal rewards, we also compared the effects
of tangiblc rewards in studies of children versus coilege students. This revealed that
9
even though tangible rewards significantly undermined intrinsic motivation for both
groups, the undermining effect was significantly greater for children than for college students on both behavioralfand self-report measures of intrinsic motivation.
The real-world implications of this patteni of results are extremwely impo(rtwat. There
is great concern about children's motivation for schoolwork, us well as for other
behaviors such as sports, art, and prosocial activities, and a study conducted by
Boggiano, Barret, Weiher, McClelland, and iusk (I987) indicated that adults tend
to view salient extrinsic rewards as an effective motivational strategy for pronioting these behaviors in children. However, the age-effect analyses indicate that,
although tangible rewards mrRay control immediate behaviors, they have negative
consequences for subsequent interest, persistence, and preference for challenge,
especially for children. Tn summa;ry, the age effects that emerged from our metaanalysis indicate that tangible rewards have a more negative effect on children than
on college students and that verbal rewards have a less positive eilcct on children
than on college students.
Unexjpected Rewards and T'ask-NoncontingentRewvards
tested
the CET prediction that unexpec ted rewards would not be detriWe next
mental to intrinsic motivation, whereas expected rewards would. The reasoning
was that if people are not doing a task in order to get a reward, they are not likely
to experience their task behavior as being controlled by the reward. The metaanalysis supported the hypothesis. Nine studies of free-choice behavior revealed
no underininig (d = 0.01 Cl = ---. 20, 0.22), and five studies of self-reported interest revealed sinilar results (d= 0.05; Cl = -().19, 0.29).
In contrast, analyses of expected rewards did yield undermining for both freechoice behavior (k -- 92: d = --0.36; Ci = -0.42, -0.30) and self-reported interest
(k = 69; d = -0.07; Cl = -0.13, -0.011). It is interesting in this regard to note that
verbal rewards are generally unexpected, and that may be one of the reasons they
do not typically have a negative efiect on intrinsic motivation.
According to CET, rewards not reqruiring task engagement should be unlikely
to affect intrinsic motivation for thie task because the rewards are not given for
doing the task. AlthoLgh relatively few studies of task-noncontingent rewards have
been done, the mteta-analysis revealed no evidence that these rewards significantly
affected either measure of intrinsic motivation (k =7; d - -0.14; Cl = -0.39, 0.11,
for free-choice behavior and k = 5; d = 0.21; Ct = -008, 0.50. for self-reported
interest).
Engagenient-ContringentRewards
E-ngagement-contingent rewar(ds are offered explicitly for engaging in an activity. Whien children were told they would get a good player award for working on an
art activity (Lepper et al., 1973), the reward was engagemniet contingent. S'imilarly
when college students were told they would receive a reward if they perforned a
hidden-figures activity, the reward was engagement contingent (Ryan et al., 1983).
lIn neither case was there a perfotnnance requiremoent: Participants did not have to
finish the task or do well on it; they simply had to work on it. More studies have
used engagement-contingent rewards uhan any other reward contingency, and that
is particularly true for studies of children. Results of the meta-analyses confinned
that engagenment-contingent rewards significanly diminished intrinsic motivation
10
In the first taxonomy of reward contingencies, Ryan et aL (1983) included taskcontingent rewards, and Cameron and Pierce included the category in their metaanalysis. Because thle task-contingent reward category is simply the aggregate of
engagemiienit-contingent re-wards and completion-contingent rewards, this category is
redundant. lowever, for nomparative purposes, we mention it here. rask-contingent
rewards undenrined intrinsic motivation assessed with both measures (k = 74: d =
-(.39; Cl = -0.46, ).32, for free choice and k - 48; d= -0.12; CI = -0.20, -0.04,
tor self-reports). Again, the undermining tended to be worse for children.
Performance-Con tingent Rewanrls
Fromn the standpoint of CET, performance-contingent rewards are the most
interesting type of tangible rewards. Perfomiance-contingent rewards were defined
by Ryan et al. (1983) as rewards given explicitly for doing well at a task or for performing up to a sDecified standard. Exampies of performance-contingency studies
include the Ryani et al. study, in which alt participants in the p.erformance-coningentrewards condition received $3 for "having donie well at the activity," and the
Harackiewicz, Manderlink, and Sansone (1984) study, in which participants
received a reward because thev were said to have performed better than 80% of
other participants.
1l1
the participants that they had not performed well. These participarnts got less than
the maximumnavailable rewards. thus in(dicating that their competence was not
optimal. For example, in a study conducted by Rosenfield, Folger, and Adelman
(1980) that involved a feedback control grouip, rewarded participants got a small
reward for performing in the bottom 15% of all participants, anid the corresponding
control group received thc comparable "negative" feedback without the reward.
Clearly, this and other such studies are quite different from the more typical studies of performance-contingent rewards in which all participants receive the same
maximum reward for having done well.
Studies involving different types of control groups and different levels of performaince were aggregated without commient by Cameron and Pieerce (I1994). in our
meta-analysis, however, because perfonnance-contingeilu rewar d effects were not
homogeneous, we examined four categories of performance-contingent rewards
rather than simply discarding outliers as Cameron and Pierce had done. The four
categories were as ifolows: effects involving no-feedback control groups in which
everyone received the maximum possible rewards, effects involving no-feedback
control groups in whlich all participants did not receive the maximum possible
rewards, effects involving comparable-feediback control groups in which all participanits received positive feedback, and effects involving comparable feedback
control groups in which all participants received negative feedback.
With the free-choice measure, for studies that compared no-feedback control
groups and participants who received the maximum possible rewards, there was
significant undennining (k = 18; d -- 0.15; Cl = -0.3 1, -0.00).' For studies with nofeedback control groups in which all participants did not receive the laximum possible rewards, there was ailso significant undenrining (k = 6; d =--0.88; CI -1.12,
-0.65). The same was tmre for studies with comparable-feedback control groups in
which everyone received positive feedback (k = 10; d = --0.20; Cf = -- )37, -0.03).
However, for the three studies with comlparable-feedback control groups invwhich
participants received negative feedback, there was not a significant effect for reward
versus no reward.
The group in which at least some participants got less than the maximumn possible rewards and the control group received no feedback stands out and deserves
special mention. This r epresents the type of perfonnance-colitingent rewards th at
one would typically find in the real world, in that here rewards are a direct function of performance. Those who perform best get the largest rewards, and those
who perfonn less well get smaller rewards or no rewards. The analysis showed that
this type of reward had the largest undermining effect of any category used in the
entire meta-analysis (d = -0.88), indicating clearly that rewarding people as a
direct function of performance runs a very serious risk of negatively affecting their
intrinsic mrotivation.
Summary of ihe PrimaryAnalyses
To surmmarize dte primarv findings from the meta-analyses, when free-choice
behavior was used as the dependent measure, all rewards, al tangible rewards, all
expected rewards, engagement-contingent rewards, completion-contingent rewards,
task-contingent rewards, and performance-contingent rewards significantly undmernmined intrinsic miotivation. Only verbal rewards enhanced intrinsic motivation in
general, but verbal rewards (lid undennine intinlsic motivation if they were given
13
if a task is dull and boring. the issue is not whether the rewards will lead people to
find the task intrinsically interesting because rewards do not add interest value to
the task itself. Rather, the issue is how to facilitate people's understanding the
importance of the activity to themselves and thus intemalizing its regulation so
they will be self-motivated to perfortmi it.
Sunumary and Conclusions
'.o sunmnarize, results of the nmeta-analysis make clear that the undenmiining of
intrinsic motivation by tangible rewards is indeed a significarnt issue. Whereas verbal rewards tended to enhlance intrinsic motivation (although not for children and
not when the rewards were given coitrollingly) and neither unexpected tangible
rewards nor task-noncontingent tangible rewards affected intrinsic motivation,
expected tangible rewards did significantly and substantially undermine intrinsic
motivation, and this effect was quite robust. Furthermore, the 1Udermining was especially strong for children. Tangible rewards-both material rewards, such as pizza
patlies for reading books, and symbolic rewards, such as good stLdent. awards- --are
wideiy advocated by many educators and are used in many classrooms, yet the
evidence suggests that these rewards tend to undermine intrinisic motivation for
the rewarded activity. Because the uindermining of intrinsic motivation by tangible rewards was especially strong for school-aged children, and because studies have
linked intrinsic motivation to high-quality leaming and adjustment (e.g., Benware
& Deci, 1984; Ryan & Grolnick, 1986), tlie findings fromii this mueta-analysis are of
particular import for primary and secondary school educators.
Specifically, the results indicate that, rather thanl focusing on rewards for motivating students' learning, it is inmpotant to focus more on how to facilitate intrinsic motivation, for example, by beginning from the students' perspective to develop more
interesting leaming activities, to provi(le mnore choice, and to ensure that tasks are
optimally challenging (e.g., Cordova & Lepper, 1 96; Deci, Sclwartz, et al., 1981:
Harter, 1974; Reeve, Bolt, & Cai, 1999; Ryan & CGrolnick, 1986: Zuckerman, Porac,
Lathin, Smith, & Deci, 1978). inthese ways, we will be more able to facilitate the type
of motivation that has been found to promnote creative task engagement (Amabile,
1982), cogritive flexibility (McGraw & McCullers, 1979), and conceptual understanding of learming activitles (Benware & Deci, 1984; Grolnick & Ryan, 1987).
I he results of fle ineta-analysis also provided strong support ror CET. Specifically, the predictions made by GET, based on an analysis of whether reward types
and reward contingencies are likely to be experienced as informational or controlling, were iuiformly supported and were particularly strong for the behavioral
measure. Thus, althouglh Cameroni and Pierce argued that CET should be abandoned and stated that there is no reason for teachers to resist using rewards in the
classroom, it is clear that CET provides an excellent account of reward effects and
that there is, in fact, good reason for teachers to think carefully about when and
how to use rewards in the classroom.
Appendix
by whether the dependent measure was free-chokce behavior or self reported interest. (Codes appear in Notes to the Appendix.)
linally,
ment ef the siusdy and results differed fromi Camieron and "ierce's If a study was
coded die same, the samne control groups were used in the comparisons, and the
effect sizes we reported did not differ fronm the effect sizes Camtieron and Pierce
reported by more than 0. 10 in eithler direction, we noted that the study was the same
in the two iet.a-aia[yses. If there was a difference, we explained what it was.
'I'able la
and Pie,-cc,(1994)
Studies used ih our meta-analysescomipared withs C(amemno
Study
Variables
P, ],1<5
v, 2, S
Arkes, 1979
Amold, 1976
Arnold, 1985
Bartelne, 1983. (D)
Blmnck et 2l., 1984, Exp. I
C, 2, F, S
E, 2, S
E C', , S
P, 2,8S
VY2, F, S
V,2.F,S
E, P, 1,
E. 1, F
F. CI , , P'
1.6
fs; 2,S.
VF. iF
Variables
E, 2, F
E,P, I,FS
C, 2, F, S
Chimng, 1995
Coohen, 1974 (D)
Crino& White, 1982
Dafoe, 1985 (D)
Dtniel &Esse., 1980
E,P,D, I,F
V,P,2,PF,S
V,2, F,S
N, P, I, F, S
P,D,2, F S
V. 1,5
C,D,2,S
I',E, I, FiS
C. 2, F, S
V 2. F, S
N, 2, F
Same.
Same.
continued
17
Earn, 1982
Variables
V, C' 2, F
-V,2, '
k, l,F
E, l,t S
1V P, I,F,S
\, 2, F,S
V, E,P,2, S
U C,qD, 1,F
P, 2, F
C,P, 1,F
C, 1,F
Fabeset.id., 1986
Fabes et al., 1988
E, 1,F, s
E,1, F, S
E, I F
C, 2, FI
V, C, P,1, F, S
C,2, F
lI,E,P, l.i
E L), l,
18
Variables
C, .F
E, C D. 2, S
Haxackiewicz, 1979
V,l?,P, 1,5
P, 1,S
P, 2, b, s
If, , 2, F, S
P, 2, F, S
P,t,S
E,D,2, F,
E, P, I, F
E, P,I.,F
V, IC ,is
V, 2, F, S
V,l,S
U,1,S
C,l,S
Table 1a
(continued)
Variables
Study
P, 2, S
U, E, I, F
Exp. 3
I (D)
2 (D)
i9)79
E, L '
(, 2,F,S
C, 2 , S
f. D, 1, F
C, P,2,FP,S
C', 2, S
k I,F
,
E, I. F, S
F, 1, F, S
E. 1,F,s
E,D, 1,F
E, D, 1,F
E, I,F
E, I, F, S
.V, E
P, S
V, U, P, 1. F
20
Variables
V, U. P, I, F
E, P, I, F, S
E, 1,F, S
1: P,2, F, S
p, 2, F, S
Pithmaa
et a., 1980
V',IC, 2, F
N, L, I, F
E, I, F
C, 2, F
V, U, E, 2. F, S
U E, 2, F, S
E, 1, F
P, 2, F, S
21
Stiluy
Variabes
E, l,:,S
k, 1, F, S
N, E, I,F
Ryan, 1982
IC, 2, F
V, E,',IC,2,F,s
Salancik, 1975
P, 2, F, S
Sansone, 1986
Sansone, 1989
Sansone et al., 1989
Sarafino, 1984
Shantb, 1981
Shiffian-Knutfmran 1990(D)
V, 2, S
17,2, S
V,2,S
Ek I, F, S
V,2, F, S
k, 1, 1, 1, S
22
(1994) analysis
negative feedback. C. &P. replrted a
verbal effect foir positive versus negative feedback, and then they collapsed across feedback to examine
tangible-reward effects. We did a
moderator analysis of rewards signiify
ing positive versus negative feedback.
C. &P. listed a perlbnnancecontingent self repon d = 2.80, but the
cotect d was 0.22. For ftee-choice,
there was a modest disepancy.
Same for free-chloice; they did int
include self-report.
Nerly the same for free-cloice; they
diui not include self-report.
Same for engagement contingent. In the
other group, children were rewarded
"for waiting," which is task noncontingent, but C .&P. coded it engagemient coitingeiit.
We included this study only in the supplemental meta-analysis of Riformational versus Controllitig verbal
rewards. C. & P. excluded it.
Same on verbal and engagemzent condngent. There were two perlormance-contingent groups, (ne
informational and one controlling.
There were three no-reward control
groups, one with informational positive feedback, one with controlling
positive ftedhack, and one witlt nofeedback. We compared performance-contingent bothi to
comparable-feedback controls
and no- feedback controls in the
moderator analyses. C. & P. did
only the comparable-feedback comparisons. Also, C. &P. did not do
an informationai-controlling
co mpaison.
Sanme coding. C. &P. collapscd across
positive and negative feedback conditions, but we did a moderator
mnalysis for positive versus negative.
Same.
Same.
Same.
Salne.
Same.
Excluded, type 1.For comparability
with other studies, we used only data
fromn the I -day assessmiients.
Variables
V, tJ,P, 2, F, S
r, D, 1, F
Excluded, type 1.
Excluded, type 1.In this studv, there
was also a condition called1positive
feedback, bit dic statements were
P, 2, F, S
('F, , i
(:2, S
N, F, 1, F
E, lF'
P, 2, S
E. 2.F'
V,E,2,F,S
F, P, 2, F; S
Vallerand, 1983
Valleramd &Reid, 1984
Vasta&Stirpe, 1979
V,l,S
P, 2,S
'
,2,S
C, I,F
23
Table Ia ocominued)
Variables
Study
Weimer, 1980
C, 2, F, S
E,
E, 2, 1, S
V. I, P
Note.
,2r, s
1, F, S
E, D, 2, F S
N, C 2, S
Same.
Notes
'Thc value k reps esents the tnmber of effects considered in calculating a composite
effect size. 13ecause, for any given calculation, the data were aggregated across all relevant
condidons witlin a study in order to enstive independenkce of effect sizes, k also represents
24
References
Amabilc T'. M. (1982)_ Social psychology of creativity: A conseusual assessment tecinique. Jioumal of Personalityand Social Psychology, 43, 997-1013.
Amnabile, T'. M., DeJong, W., & Lepper, M. R. (1976). Effects of externally imposed
deadlines on subsequent intrinsic motivation. Journal of Personality and Social
Psychology, 34, 92 98.
Bernware. C., & Deci, E. L. (1984). Quality of learninig with an active versus passive
motivational set. Aimeiican EducationalResearch Journal,21, 755--765.
Boggiano.A. K., Barrett. M., Weiher, A. W., McClelland, G. 1., & Lusk, C. M. (1987).
Use of the maximal-operantprincipletto iiotivate children's intrinsic interest. Jolrnal
of Personality and Social Psychology, 53, 866 -879.
Cam.eron, J., & Pierce, W. D. (i994). Reinforcement, reward, and intrinsic motivation:
A *neta-analysis. Review of EducationalResearch, 64, 363-423.
Caneron, J., & Pierce, W. 13 (1996). The debate about rewards and itrinsic motivation:
Protests and accusations do not alter the results. Review of EducationalResear.h, 66,
39-52.
Cordova, D. 1., & Lepper, M. R. (1996). Intrinsic motivation and the process of learning: Beneficial effects of contextualization, personalization, and choice. Journal of
EducationalPsychology, 88, 715-730.
deChanns, R. (1968). Personalcausation. New York: Academic Press.
D)eci, E. L. (1971). Effects of externally mediated rewards on intrinsic motivation.
Journalof Personalityatun Social Psychology, 18, 105--I 15.
Deci, E. L. (1972a). Effects of contingent and non-contingent rewards and controls on
intrinsic motivation. Organizational Behavior andHunuan Pe4brntance, 8. 217-229.
Deci, E. L. (1972b). Intrinsic motivation, extrinsic reinforcement, and inequity. Journal of Personality and Socxial Psychology, 22, 113-120.
Deci, e. L., Betley, G., Kahle; J., Abranis, L., & Porac, J. t1981). When trying to win:
Comiipetition mid intrinsic notvation. Personalityand Social Psychology Bulletin,
7, 79-83.
Deci, E. L., Connell. J. P., & Ryan, R. M. (1989). Self-determination in a work organization.Journal of Apphed Psy.chology, 74, 580--590.
Deci, E. L., Koestelr, R., & Ryan, R. M. (1999). A nieta-anualytic review of experiments
examining the effects of extrinsic rewards on intrinsic motivation. Psychological
Bulletin, 125, 627-668.
Deci, F. L., & Ryan, R. M. (1980). The empirical exploration of intrinsic motivational
processes. In L. Berkowitz (Ed.), Advances in experimental social psyc hology
(Vol. 13, pp. 39-80). New York: Academic Press.
Deci, E. L., & Ryan, R. M. (1985). Intrinsicmnotivation and self:detenrinatmn in human
behavior. New York: Plenumi.
Deci, E. L., & Ryan, R. M. (1991). A motivational approach to self: Integration ill personality. In R. Dienstbier (Ed.), Nebraska Symposium on Motivation: Vol. 38. Perspectives on motivation (pp. 237- 288). Lincoln: University of Nebraska Press.
25
LEtrinsicRewazrdl
and IntrinsicMotivation
Ryan, R. M., Miniis, V., & Koestner, R. (1983). Relation of reward contingency and
interpersonal context to intrinsic rnotivation: A review and test using cognitive evalnation theory. Journalof Personalityand SocialPsychology, 45, 736 750.
Ryan, R. M., & Stiller, 1. (1991). The social contexts of internalization: Parent and
teacher influences on autonomy, motivation and learning. In M. L. Maehr & P. R.
Pintrich (Eds.), Advances in motivation and achievement (Vol. 7, pp. 115 150).
Greenwich, CT: JAI Press.
Smith, W. E. (1975). The ejfrct of anticipatedvs. unanticipatedsocial rewardon subsequent intrinsic motivation. Unpublished (doctoral dissertation, Comnell University,
Ithaca, NY.
Zuckerman, M., Porac, J., Lathin. D., Smith, R., & Dec., E. L. (1978). On the importance of self-determination lor intrinsically motivated behavior. Personality and
Social Psychology Bulletin, 4, 443-446.
Authors
EDWARD L. DECI is Professor of Psychology, Department of Clinical and Social Sciences
in Psychology, University of Rochester, Rochester, NY 14627; deci@psych.rochester.edu.
His research examiines the effects of social contexts on motivation and self -determination
as they relate to effective functioning in various real-world domains.
RICHARD KOESTNER is Associate Professor of Psychology. McGill University, 1205 Dr.
Penfield Avenue, Montreal, Quebec H3A IB1, Canada; koestner@hebb.psycl.nmcgill.ca.
He is a clinical psychologist wvhose research focuses on factors affecting the development
and maintenance of motivation and self-regulation in personality.
RICHARD M. RYAN is Professor of Psychology, Department of Clinical aid Social Sciences
in Psychology, University of Rochester, Rochester, NY 14627; ryanpsych.rochester.edtu.
A clinical psychologist, he specializes in motivation, self-determination, and well-beinlg,
and their application to numenrous areas including the impact of educational reform.
27
COPYRIGHT INFORMATION